Article URL: https://quickchat.ai/post/expose-ai-agent-as-mcp-server Comments URL: https://news.ycombinator.com/item?id=48797219 Points: 5 # Comments: 0
Article URL: https://www.wired.com/story/fact-checking-ai/ Comments URL: https://news.ycombinator.com/item?id=48797126 Points: 1 # Comments: 0
Article URL: https://github.com/momomuchu/make-no-mistakes Comments URL: https://news.ycombinator.com/item?id=48797101 Points: 1 # Comments: 0
Article URL: https://arpitbhayani.me/blogs/ai-topological-sort/ Comments URL: https://news.ycombinator.com/item?id=48797087 Points: 1 # Comments: 0
<p>Exclusive: Foreign secretary warns of combined risks of AI, climate crisis, irregular migration and foreign interference</p><p>Artificial intelligence poses a “Hiroshima”-style risk to humanity if governments do not agree to curb how it is developed, the foreign secretary has warned.</p><p>Yvette Cooper urged countries, including the US and China, to agree international rules for AI, telling the Guardian she believes the issue will dominate foreign policy over the next two years.</p> <a href=
Article URL: https://intextbooks.science.uu.nl/workshop2026/files/itb26_s1s2.pdf Comments URL: https://news.ycombinator.com/item?id=48796817 Points: 19 # Comments: 3
Article URL: https://www.nomlings.cc/ Comments URL: https://news.ycombinator.com/item?id=48796800 Points: 4 # Comments: 1
Article URL: https://www.tomshardware.com/tech-industry/atomic-semi-rebrands-as-fab2-and-shifts-operations-to-texas Comments URL: https://news.ycombinator.com/item?id=48796784 Points: 11 # Comments: 0
The effect renders a 3D avatar model on top of the user's face. Tracks one face at a time. It uses MediaPipe's GPU delegate (WebGL2) when available, with automatic fallback to a WASM CPU delegate. MediaPipe's FaceLandmarker runs inside a Web Worker that the SDK spawns. The worker dynamically imports MediaPipe, its WASM, and the 3D model. The avatar requirements: a 3D mesh with Apple ARKit Facial Blendshapes in GLB file format. I welcome everyone who tries it and posts a feedback here! Comments U
The main metric I look at is inference cost, which is a pretty good indicator when something's off. However, a couple of days a go I discovered we were loading ~300 tools due to a bug, affecting both costs and accuracy. It was going on for some time and I didn't realize it by looking at the costs, cause it got hidden by other changes I made in the same period. Now I started tracking that and some other metrics, but it made me wonder what else I might be missing. So: what are you tracking? Any ti
Article URL: https://gamefork.io Comments URL: https://news.ycombinator.com/item?id=48796620 Points: 2 # Comments: 0
<pre style='white-space:pre-wrap;width:81ex'>fix: land ten small reliability fixes (#100399) * fix(cron): reject sub-millisecond durations * fix(skill-workshop): preserve proposal terminal newline Preserve proposal_content exactly at the agent tool boundary and make renderProposalMarkdown defensively emit a terminal newline. Add focused regressions for the tool write path and markdown renderer. * fix(skill-workshop): reject blank raw proposal content * fix: treat empty-string optional integer to
With some corporate environments going from "tokenmaxxing at all costs" to now setting strict token budgets for agentic development: How is one supposed to adhere to a managerial / leadership instruction of "use AI for everything" but even more than that: "you shouldn't be writing code by hand"? It feels like this is a Russian roulette scenario. As long as you have the right harness, framework, skills, prompt, etc. to make the output as deterministic as possible AND know exactly what you're supp
<!-- SC_OFF --><div class="md"><p>I'm an 18-year-old independent student from Tunisia. I built and I'm leading an open, from-scratch machine-translation pipeline and parallel corpus for Tunisian Darija. Sharing it for feedback.</p> <p>Why: Tunisian Darija, written in Arabizi (Latin letters + numerals like 3/7/9/5 for Arabic phonemes), has almost no open NLP resources. Existing Arabic tools route it through MSA and mishandle the orthography. To the best of my knowledge there was no open parallel
Article URL: https://github.com/vaishcodescape/shipd-agent Comments URL: https://news.ycombinator.com/item?id=48796341 Points: 1 # Comments: 0
Article URL: https://github.com/salarkhannn/ghostlog Comments URL: https://news.ycombinator.com/item?id=48796330 Points: 1 # Comments: 0
<h2> El problema de los pipelines "texto → SQL" tradicionales </h2> <p>La mayoría de las soluciones de IA para bases de datos siguen el mismo patrón: el usuario escribe una pregunta en lenguaje natural, un modelo la traduce a una consulta SQL, y esa consulta se ejecuta directamente contra la base de datos. Es simple, pero frágil: si el modelo genera una consulta incorrecta, esta puede <strong>ejecutarse sin errores</strong> y devolver un resultado que parece válido pero no lo es. Nadie se entera
<p>I built <a href="https://vibesafe.store" rel="noopener noreferrer">VibeSafe</a> — a 24-check security scanner for apps made with AI coding tools (Cursor, Lovable, Bolt, v0, Replit).</p> <p>The idea came from a pattern I kept seeing: AI tools produce code that <em>works</em>, but rarely code that's <em>secure</em>. Hardcoded API keys, disabled Row Level Security, missing CSP headers, exposed <code>.env</code> files. Same issues, different apps.</p> <p>Before writing a line of scanner code, I m
<h2> Introduction </h2> <p>One of the biggest limitations of modern AI assistants is memory. Every new conversation starts almost from scratch, making it difficult to work on long-running investigations or complex cases.</p> <p>For my hackathon project, I wanted to solve exactly this problem.</p> <p>I built <strong>DetectiveAI</strong>, an AI-powered crime investigation assistant that uses <strong>Cognee Cloud</strong> as its persistent memory layer. Instead of forgetting evidence after every se
Article URL: https://github.com/HBarefoot/engram Comments URL: https://news.ycombinator.com/item?id=48796133 Points: 2 # Comments: 0
Article URL: https://istota.cynium.com/ Comments URL: https://news.ycombinator.com/item?id=48796095 Points: 5 # Comments: 0
Article URL: https://github.com/adam-s/goldseam Comments URL: https://news.ycombinator.com/item?id=48796042 Points: 1 # Comments: 0
Article URL: https://github.com/guy-lifshitz/tacheles Comments URL: https://news.ycombinator.com/item?id=48796026 Points: 1 # Comments: 0
Article URL: https://allk12.com/blog/why-high-earning-families-are-leaving-traditional-schools-for-ai Comments URL: https://news.ycombinator.com/item?id=48796022 Points: 2 # Comments: 0
I'm a father of two, 7 and 12. They are obsessed with Roblox, especially Rivals. Like a lot of parents, we did not love it. We tried the usual things: block it, limit it, set timers." It became a daily battle, a lose-lose situation. So I flipped the problem. Instead of fighting what they loved, I decided to lean into it, but with a twist. Why just play an FPS when you could build one together? My kids became the PMs. Claude and I became their engineer. I was shocked by how fast we moved. We pick
<pre style='white-space:pre-wrap;width:81ex'>docs: tailor imsg skill for OpenClaw agents (#100105) * docs: tailor imsg skill for OpenClaw agents * docs(imsg-skill): document that poll send echoes --question as a visible comment * docs(imsg-skill): note inbound native poll question is backfilled from its caption * docs(imsg-skill): document only shipped poll behavior Drop the poll caption echo, --comment override, and inbound question backfill from the imsg skill: those live in imsg PR openclaw/i
<pre style='white-space:pre-wrap;width:81ex'>test: avoid redundant prompt probe process</pre>
<pre style='white-space:pre-wrap;width:81ex'>fix(test): unit-fast tests inherit live host config (#100221) * fix(test): make unit tests hermetic against host config and build state Two host-state leaks made the full suite fail deterministically on developer and agent machines while CI stayed green: - Built checkouts: bundled-plugin manifest discovery prefers dist/extensions, which by design excludes externalized official plugins (e.g. qwen). Tests that depend on manifest-driven endpoint classifi
Article URL: https://arxiv.org/abs/2602.09239 Comments URL: https://news.ycombinator.com/item?id=48795830 Points: 73 # Comments: 42
Article URL: https://www.lesswrong.com/posts/zJGGZQdtfoNye5ywe/probing-the-loss-band-sparsity-assumption-in-scientist-ai Comments URL: https://news.ycombinator.com/item?id=48795793 Points: 1 # Comments: 0
Article URL: https://github.com/wonsukchoi/domain-experts Comments URL: https://news.ycombinator.com/item?id=48795717 Points: 2 # Comments: 3
Article URL: https://github.com/rbsriram/heckle Comments URL: https://news.ycombinator.com/item?id=48795580 Points: 3 # Comments: 3
<pre style='white-space:pre-wrap;width:81ex'>fix(clickclack): gate provenance stamping behind agentActivity opt-in</pre>
<pre style='white-space:pre-wrap;width:81ex'>fix(clickclack): forward native activity callbacks Allow ClickClack-owned progress rendering to receive item callbacks when source delivery is suppressed. This lets native agentActivity persist commentary and tool rows instead of relying on default progress texts. HyperReview-Reflex: pass HyperReview-Tier: light HyperReview-Scope: staged-tree:e4d63a7648ad03f91fe81588cce783fd686b6488 HyperReview-Paths-SHA256: 94ae4829752bfdafc02191354cac5ae2a2ca9fc899a
<pre style='white-space:pre-wrap;width:81ex'>feat(clickclack): stamp model/thinking attribution onto agent posts Wires replyOptions.onModelSelected so the resolved provider/model and thinking level for each turn (including after fallback) are sent as author_model / author_thinking fields on activity rows and the final reply. Servers without these columns ignore the unknown JSON fields, so the wire shape is backward compatible; servers that persist them get per-message attribution.</pre>
<pre style='white-space:pre-wrap;width:81ex'>docs: regenerate docs_map for clickclack agent activity heading</pre>
Article URL: https://meteoraweb.com/en/news/ai-powered-hyper-communication-leads-277-americans-to-rank-internet-medicine-and-democracy-as-top-us-contributions-for-250th-birthday Comments URL: https://news.ycombinator.com/item?id=48795513 Points: 1 # Comments: 0
Article URL: https://maxghenis.com/drinking-ai/ Comments URL: https://news.ycombinator.com/item?id=48795424 Points: 1 # Comments: 0
We needed hyper-lightweight sandboxes for massively parallel agent fanout that could be recycled in milliseconds, take only KBs of memory when not actively running a command, and can work with bottomless filesystems. Tinysandbox is like just-bash, but supports a minimal secure JS runtime via QuickJS WASM, random file IO, VFS checkpoint and restore, and streaming IO for working with large files. It's designed for super-lightweight agents that don't need full linux, but still need the emergent beh
A Google Deepmind developer ported the 2003 real-time strategy game "Command & Conquer: Generals Zero Hour" to iPhone and iPad using Anthropic's Claude Code. The first build took 40 minutes. The full source code is on GitHub. The article Claude Code and Fable 5 ported the 2003 PC game Command & Conquer to native iOS in "a few hours" appeared first on The Decoder .
